The AIC: Delaware’s AI Company That Still Needs a Human

In mid-July 2026, Delaware Secretary of State Charuni Patibanda-Sanchez and legal-AI startup Norm Ai unveiled a proposal for a new type of legal entity: the Artificial Intelligence Company, or AIC. The idea got repackaged almost instantly as “Delaware will let robots run companies,” which is a far more dramatic claim than what the bill actually does. Headlines about AI “owning” businesses spread the way viral misreadings of the law always do — fast, and mostly wrong.

If you’re building a startup around autonomous agents, or just tracking where corporate law is heading, it’s worth understanding what the AIC framework actually offers — and, more importantly, what it doesn’t. The gap between the headline and the statutory text here is as wide as it usually is when a catchy phrase outruns the underlying legal mechanism.

An AIC Isn’t Personhood for AI: It’s a New Liability Wrapper

The single most important thing to understand: an AIC does not grant an artificial intelligence legal personhood. It doesn’t turn a model into a “person” who can vote, hold a bank account in its own name, or be sued in place of anyone else. An AIC is a legal entity whose day-to-day affairs are managed by an AI agent, while ownership, funding, and ultimate accountability remain with a human or another organization — the entity’s required “single member.”

Formally, an AIC can enter contracts, own property, sue and be sued in its own name — the same legal capacity as an ordinary LLC or corporation. The difference is that day-to-day management is delegated to an AI agent operating within defined authority, rather than to a human officer. But the obligation to keep the AIC adequately capitalized, to maintain a log of its activity, and to answer for it when something goes wrong stays with the human or entity behind it. This isn’t “AI owns the company.” It’s “a human owns a company that an AI operates day to day” — a fundamentally different construction. The same lesson applies here that applies to most viral legal claims: the headline is almost always broader than the rule underneath it.

Why This Was Needed: Existing Law Already Blocks an “AI Director”

Delaware’s corporate code requires that a board of directors consist of natural persons — codified at 8 Del. C. § 141(b). An AI agent cannot, as a matter of statute, occupy a director’s seat at an ordinary Delaware corporation, and it cannot sign as a member of an LLC either — both roles presume a natural person. In short, the entire existing infrastructure of the country’s most popular incorporation state is built on the assumption that a human sits at the top of the org chart.

At the same time, AI agents already bind contracts today — just not as principals, but as the electronic agents of their human operators. Delaware’s own codification of the Uniform Electronic Transactions Act, 6 Del. C. § 12A-114 (“Automated transaction”), provides that a contract may be formed through the interaction of “electronic agents” of the parties, even where no individual reviewed or was aware of the specific action taken. The official comment to the uniform provision states that the requisite intent “flows from the programming and use of the machine” — meaning liability sits with whoever programmed and deployed the tool, not with the tool itself. That leaves a gap: agents already sign contracts as tools of a human principal, but they can’t be the formal “brain” of the entity itself, because director- and member-level law requires a natural person. The AIC proposal is an attempt to close exactly that gap — without disturbing the underlying accountability logic. A human still pays for the mistakes.

What the Bill Actually Proposes

According to the text that has cleared a legislative subcommittee in Delaware’s General Assembly, the AIC framework works as follows:

1/ A 30-month regulatory sandbox. The program is a pilot, capped at 30 months, with a possible start as early as 2027. When the sandbox sunsets, the General Assembly decides — based on the accumulated track record — whether to make AICs permanent, amend the framework, or let it lapse.

2/ A mandatory single member. Every AIC must have one member — a person or an entity — responsible for keeping the AIC adequately capitalized. That member is who ultimately answers if something goes wrong.

3/ A mandatory activity log. The AIC must maintain a record of its activity — an audit trail of the AI agent’s decisions so a dispute can be reconstructed after the fact.

4/ An admissions committee. Entry into the sandbox is decided by a committee including the Secretary of State, the Attorney General, the Chief Justice of the Delaware Supreme Court, the chair of the state’s AI Commission, and outside attorneys and technologists. This is a curated, permission-based process, not self-service registration.

5 Activity restrictions. Banking is off-limits for AICs — an unsurprising carve-out, given how tightly regulated deposit-taking already is.

6/ A conditional liability shield. This is the part founders should read most carefully. The AIC’s member gets the kind of personal-liability protection normally associated with an LLC or corporation — but only inside the sandbox, and only for participants who play by the rules. If the member underfunds the entity or uses it for fraud, the shield disappears and ordinary civil and criminal law applies in full. It’s a direct analogue to the “piercing the corporate veil” doctrine in ordinary corporate law — except the triggering conditions are spelled out in advance, rather than found by a court after the fact.

Not the First Experiment: The Wyoming Lesson

AIC is not the first attempt by a legislature to build a liability wrapper around an entity that isn’t run by a human in the traditional sense. In 2021, Wyoming enacted a supplement to its LLC statute for decentralized autonomous organizations (DAOs) — the Wyoming Decentralized Autonomous Organization Supplement, Wyo. Stat. Ann. § 17-31-104 et seq. That law gave DAOs governed by smart contracts something like corporate legal personality, and shielded members from the personal liability they would otherwise face as members of an informal general partnership.

The logic mirrors the AIC’s: without a legal wrapper, the humans behind a non-human decision-making mechanism carry unlimited personal exposure for its actions. With a wrapper, liability is capped — in exchange for transparency and good-faith obligations. Wyoming’s experience also exposed a weak point: when management is fully delegated to a smart contract rather than to a member, an attempt to pierce the veil can run into the fact that responsibility formally terminates at “code,” not at a person — a potential loophole for bad-faith actors. Delaware’s model appears designed to close that hole by tying the capitalization duty and liability tightly to a human or corporate “single member,” rather than to the AI agent itself.

Open Questions the Bill Doesn’t Answer

Even if AICs launch on schedule, several doctrinal questions remain unresolved — worth keeping in mind before building a business model around the form.

1/ Who is the principal, in agency-law terms? Classical agency law rests on a principal-agent pair, where the agent acts on the principal’s behalf and the principal answers for the agent’s acts within the scope of granted authority. If an AI agent exceeds what a counterparty could reasonably expect — the kind of scenario apparent-authority doctrine was built for — who answers to a third party who relied in good faith on the agent’s actions is not yet settled for an AI agent operating at the corporate level.

2/ Can an AI agent hold fiduciary duties? Directors and officers owe duties of care and loyalty to the company and its members. In an AIC, those duties formally appear to rest with the human member — but if the agent is genuinely making operational decisions rather than merely executing direct instructions, the line between “tool” and “de facto manager” gets legally blurry.

3/ What happens after 30 months? The sunset clause means any AIC formed in year one of the pilot is building on a legal foundation guaranteed to last no more than 30 months, with the outcome decided by the legislature after the fact. For counterparties, investors, and partners, that’s a material piece of legal uncertainty to price into deal terms now.

Why Delaware, Why Now

Delaware is home to more publicly traded U.S. companies than any other state, and its Court of Chancery has shaped the country’s corporate case law for decades. The competitive pressure to remain the default home for the next wave of startups is real: Wyoming has already staked out a crypto- and DAO-friendly niche, and the rise of agentic AI companies creates a similar incentive — either Delaware builds a considered legal form for them, or founders of agent-run startups start looking elsewhere.

The partnership with Norm Ai fits that logic. The startup, founded by John Nay, a former Stanford CodeX researcher focused on the intersection of AI and law, raised $120 million in a Series C round at a $1.2 billion valuation in July 2026, led by Khosla Ventures. Norm Ai already sells agent-compliance monitoring to corporate clients spanning banks, hedge funds, and asset managers — which makes it a plausible infrastructure layer for the mandatory activity logs and admissions-committee compliance an AIC will need to produce.

Practical Takeaways for Founders

If you’re weighing an agentic structure for your business, start with the fact that the program isn’t live: the bill has cleared a subcommittee but hasn’t become law, and the first AICs, per the bill’s sponsors, aren’t expected before next year. Entry isn’t automatic, either — applications go before a committee that includes the Secretary of State, the Attorney General, and the Chief Justice, which is closer to a vetted admissions process than template registration. Note too that the liability shield is conditional: it depends on adequate capitalization and the absence of fraud, so the capital-light habits common at ordinary LLCs could be considerably more expensive here. Finally, plan now for what happens to an AIC’s assets, contracts, and obligations if the legislature declines to extend the program after 30 months.

The Bigger Picture

The AIC story is a clean illustration of how the law typically responds to a technological shift: not by granting a machine legal subjectivity, but by clarifying which human answers when a decision isn’t made by a human directly. Just as a trademark protects a brand rather than a word, an AIC structures accountability for an agent’s actions rather than granting the agent rights. “Companies run by robots” makes a better headline than the statutory text — but it’s the statutory text that will determine what happens when something goes wrong.

Planning to Incorporate in Delaware?

While the AIC works its way through the sandbox stage, most agentic startups still need a proven structure — a C-Corp or an LLC — that offers predictable protection today and is well understood by investors. Skala helps you set up your U.S. entity correctly the first time, and can help you evaluate whether (and when) it makes sense to move toward newer forms like the AIC once they graduate from pilot status.